Hypoxia facilitates Alzheimer’s disease pathogenesisby up-regulating BACE1 gene expressionXiulian Sun*†, Guiqiong He*‡, Hong Qing*, Weihui Zhou*, Frederick Dobie*†, Fang Cai*, Matthias Staufenbiel§,L. Eric Huang¶, and Weihong Song*†�

*Department of Psychiatry, Brain Research Center, and †Graduate Program in Neuroscience, University of British Columbia, 2255 Wesbrook Mall,Vancouver, BC, Canada V6T 1Z3;‡Department of Human Anatomy, Chongqing University of Medical Sciences, Chongqing 400016, China;§Nervous System Disorders Research Group, Novartis Institutes for BioMedical Research, CH-4002 Basel, Switzerland; and¶Department of Neurosurgery, University of Utah School of Medicine, Salt Lake City, UT 84112

Edited by L. L. Iversen, University of Oxford, Oxford, United Kingdom, and approved October 12, 2006 (received for review July 24, 2006)

The molecular mechanism underlying the pathogenesis of the ma-jority of cases of sporadic Alzheimer’s disease (AD) is unknown. Ahistory of stroke was found to be associated with development ofsome AD cases, especially in the presence of vascular risk factors.Reduced cerebral perfusion is a common vascular component amongAD risk factors, and hypoxia is a direct consequence of hypoperfusion.Previously we showed that expression of the �-site �-amyloid pre-cursor protein (APP) cleavage enzyme 1 (BACE1) gene BACE1 is tightlycontrolled at both the transcriptional and translational levels and thatincreased BACE1 maturation contributes to the AD pathogenesis inDown’s syndrome. Here we have identified a functional hypoxia-responsive element in the BACE1 gene promoter. Hypoxia up-regu-lated �-secretase cleavage of APP and amyloid-� protein (A�) pro-duction by increasing BACE1 gene transcription and expression bothin vitro and in vivo. Hypoxia treatment markedly increased A� dep-osition and neuritic plaque formation and potentiated the memorydeficit in Swedish mutant APP transgenic mice. Taken together, ourresults clearly demonstrate that hypoxia can facilitate AD pathogen-esis, and they provide a molecular mechanism linking vascular factorsto AD. Our study suggests that interventions to improve cerebralperfusion may benefit AD patients.

hypoxia-inducible factor 1� � amyloid-� protein � neuritic plaque �memory deficit � transcription

Deposition of amyloid-� protein (A�) in the brain is thehallmark of Alzheimer’s disease (AD) pathology (1). A�,

the major component of neuritic plaques, is derived from�-amyloid precursor protein (APP) after sequential cleavage by�- and �-secretase. Early-onset familial AD caused by mutationsin APP and in the presenilin 1 and 2 genes accounts for only�5% of total AD cases. The majority of AD cases are sporadicAD with late onset and have no defined cause. The major riskfactors for AD include aging, atherosclerosis, diabetes mellitus,stroke, the ApoE �4 polymorphism, and less education. Recentstudies have shown that a history of stroke can increase ADprevalence by �2-fold among elderly patients (2–6). The risk ishighest when stroke is concomitant with atherosclerotic vascularrisk factors (7). Patients with stroke or cerebral infarction alsoshow poorer cognitive performance and greater severity ofclinical dementia (8). Hypoxia is a direct consequence of hypo-perfusion, a common vascular component among the AD riskfactors, and may play an important role in AD pathogenesis.

Oxygen homeostasis is essential for the development and phys-iology of an organism. Hypoxia-inducible factor 1 (HIF-1) is theprincipal molecule regulating oxygen homeostasis (9). HIF-1 is amember of the basic helix–loop–helix transcription factor family,and the basic region of the protein binds specifically to the5�-RCGTG hypoxia-responsive element (HRE) in a gene promoterregion. HIF-1 contains an oxygen-regulated expression subunit �(HIF-1�) and a constitutively expressed subunit � (HIF-1�) (Arnt).HIF-1� protein, mediated by its oxygen-dependent degradationdomain, is rapidly degraded through the ubiquitin–proteasome

pathway under normoxic conditions with a half-life of �5 min, butis quite stable under hypoxic conditions (10). The hypoxia signaltransduction pathway plays a major role in vascular developmentand ischemia, as well as in neurodegeneration (11, 12). Althoughthe brain accounts for only 2% of the body’s mass, it utilizes �20%of total resting oxygen and is thus particularly susceptible toconditions of hypoxia. Neurons in the hippocampus and neocortexwere shown to be selectively affected by cerebral ischemia (13, 14).When oxygen is in short supply, HIF-1 binds to HRE in promotersor enhancers, thereby activating a broad range of genes involved inangiogenesis, erythropoiesis, cell death, and energy metabolism(15). HIF-1 is up-regulated in the human frontal cortex with aging(16). Activation of the HIF-1 pathway by risk factors such as stroke,age, and cerebral vascular atherosclerosis may contribute to ADpathogenesis by facilitating A� deposition.

�-Site APP cleavage enzyme 1 (BACE1) is the �-secretase invivo, and BACE2 is a homolog of BACE1. Even though BACE1and BACE2 are highly homologous (17), they have distincttranscriptional regulation and function. BACE2 is not a �-secre-tase but rather functions as a novel �-secretase to cleave APPwithin the A� domain (17, 18). In addition to processing APP atthe �-secretase site, BACE1 was found to cleave other proteins(19–23). BACE1 interacts with reticulon family member pro-teins, and reticulon proteins block access of BACE1 to APP andreduce APP cleavage (24). BACE1 gene expression is tightlycontrolled at both the transcriptional and translational levels (17,25–27). BACE1 protein and activity levels increase with agingand in some AD brains (28–32). Our recent finding suggests thatabnormal BACE1 trafficking and maturation contribute to theAD pathogenesis in Down’s syndrome (33). These studies indi-cate that up-regulated BACE1 gene expression at the level oftranscription or translation could contribute to AD pathogenesisin some sporadic cases.

In the study described here, we found that hypoxia signifi-cantly increased BACE1 gene expression, resulting in increased�-secretase activity and A� production. Furthermore, hypoxiatreatment markedly increased A� deposition and potentiatedthe memory deficit in AD transgenic mice. Our results demon-strate that hypoxia can facilitate AD pathogenesis, and theyprovide a molecular mechanism to link vascular factors to AD.

Author contributions: G.H., H.Q., and W.Z. contributed equally to this work; W.S. designedresearch; X.S., G.H., H.Q., W.Z., F.D., F.C., and W.S. performed research; M.S., L.E.H., andW.S. contributed new reagents/analytic tools; X.S. and W.S. analyzed data; and X.S. andW.S. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS direct submission.

Abbreviations: A�, amyloid-� protein; AD, Alzheimer’s disease; APP, �-amyloid precursorprotein; BACE1, �-site APP cleavage enzyme 1; HIF, hypoxia-inducible factor; HRE, hypoxia-responsive element.

�To whom correspondence should be addressed. E-mail: weihong@interchange.ubc.ca.

© 2006 by The National Academy of Sciences of the USA

www.pnas.org�cgi�doi�10.1073�pnas.0606298103 PNAS � December 5, 2006 � vol. 103 � no. 49 � 18727–18732

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Nov

embe

r 11

, 202

1

Results and DiscussionHypoxia Up-Regulates BACE1 Promoter Activity. Previously, wecloned the human BACE1 gene promoter and found that BACE1gene expression is tightly controlled at both the transcriptional andtranslational levels (26, 27). Here, a series of deletion luciferasereporter plasmids containing human BACE1 gene promoter re-gions were constructed to investigate whether BACE1 is one of thedownstream target genes of the hypoxia signaling pathway. BACE1promoter constructs containing base pairs �932 to �292 (pB1P-H)and base pairs �896 to �292 (pB1P-I) were transfected intoSH-SY5Y cells. The transfected cells were treated with 2% O2 ina hypoxia incubator for 48 h. Luciferase activity was measured asan indication of promoter activity (Fig. 1A). Hypoxia treatmentmarkedly increased the promoter activity of pB1P-H by (mean �SEM) 296.80 � 13.25% (P � 0.0001 relative to control) but had nosignificant effect on pB1P-I (P � 0.05). Consistent with previousstudies (34), the promoter activity of positive control plasmidpEpoE-Luc was also significantly up-regulated under hypoxia treat-ment (P � 0.001), and hypoxia had no effect on the vector control(Fig. 1A). Western blot analysis showed that HIF-1� was signifi-cantly induced under hypoxic conditions (Fig. 1B). These resultsindicate that the region between base pairs �932 and �896 of theBACE1 promoter may contain a hypoxia-inducible enhancerelement.

A Functional HRE Site in the Promoter Regulates BACE1 Gene Expres-sion. Hypoxia up-regulates gene transcription by stabilizing tran-scription factor HIF-1, which binds to HRE in the promoterregion of hypoxia-regulated downstream target genes. TheHIF-1 heterodimer binds to HRE consensus sequence 5�-RCGTG. Sequence analysis reveals that the human BACE1 genepromoter contains a putative HRE site that is located at basepairs �915 to �911 (5�-ACGTG) (Fig. 1C). A gel shift assay wasperformed to determine whether transcription factor HIF-1 canbind to the putative HRE site in the human BACE1 genepromoter. BACE1-HRE, a 20-bp double-stranded oligonucleo-tide probe (5�-gggaggccgACGTGggcg) corresponding to BACE1promoter region base pairs �924 to �907 was synthesized andend-labeled with 32P. A shifted DNA-protein complex band wasdetected after incubation of the BACE1-HRE probe with HeLanuclear extract (Fig. 1D, lane 2). The binding intensity of thisshifted band was significantly reduced by applying 50-fold molarexcess of unlabeled Epo-HRE consensus competition oligonu-cleotides, and the shifted band was completely abolished byaddition of 200-fold of unlabeled Epo-HRE consensus oligonu-cleotides (Fig. 1D, lanes 3 and 4). Preincubation of excessivemutant BACE1-HRE oligonucleotides containing the bindingsite mutations (5�-gggaggccgAAAAGggcg) with HeLa nuclearextract had no competitive effect on the BACE1-HRE shiftedband (Fig. 1D, lane 5). These results clearly demonstrate that thehuman BACE1 promoter contains a HRE site.

To determine whether this binding element biologically re-sponded to the transcription factor HIF-1 in the transcriptionalregulation of the human BACE1 gene, SH-SY5Y cells were trans-fected with BACE1 promoter plasmids pB1P-H and pB1P-I andwith HIF-1� mammalian expression plasmid. Transfection with

Fig. 1. Up-regulation of BACE1 gene transcription by hypoxia. (A) Hypoxiaincreases human BACE1 promoter activity. BACE1 promoter constructs pB1P-Hand pB1P-I were transfected into SH-SY5Y cells. Plasmid pGL3-Basic (vector)and pEpoE-Luc were used as negative and positive controls, respectively. Cellswere exposed to 2% O2 (hypoxia) or 21% O2 (control) after transfection.Luciferase assay was performed 48 h after transfection to reflect promoteractivity. *, P � 0.001 by ANOVA and Student’s t test. (B) HIF-1� expression wasrapidly induced by hypoxia in SH-SY5Y cells. Cells were treated with 2% O2 for0, 1, and 2 h and then lysed in RIPA-Doc buffer (0.15 mM NaCl/0.05 mM Tris �

HCl, pH 7.2/1% Triton X-100/1% sodium deoxycholate/0.1% SDS). HIF-1� wasdetected by rabbit polyclonal anti-HIF-1� antibody (H206). (C) Diagram showsbase pairs �980 to �900 of the human BACE1 promoter sequence (relative tothe transcription start site). A HRE consensus site 5�-RCGTG was located at basepairs �915 to �911 (capitalized and underlined). (D) The human BACE1promoter contains a HRE site. Gel shift assay was performed as described inMethods. A 32P-labeled double-stranded oligonucleotide probe, [32P]BACE1-HRE, corresponding to BACE1 promoter base pairs �924 to �907 was used asa probe. (E) Robust HIF-1� expression in HEK293 cells after pHIF-1� expressionplasmid transfection. (F) BACE1 promoter activity was increased by HIF-1�

overexpression. BACE1 promoter constructs pB1P-H, pB1P-I, and positive con-trol pEpoE-Luc were cotransfected with HIF-1� expression plasmid intoHEK293 cells. Luciferase assay was performed 48 h after transfection. HIF-1�

overexpression can significantly increase promoter activity in cells transfectedwith pB1P-H but not pB1P-I. HIF-1� overexpression also significantly increasedpEpoE-Luc promoter activity. *, P � 0.001 by ANOVA. (G) HIF-1� siRNAtransfection reduced HIF-1� expression in HEK293 cells. (H) HIF-1� siRNAinhibited hypoxia’s up-regulatory effect on the human BACE1 promoteractivity. BACE1 promoter construct pB1P-H or positive control pEpoE-Lucplasmid were cotransfected with control siRNA or HIF-1� siRNAs. The trans-fected cells were then exposed to 2% O2 (hypoxia) or 21% O2 (control) 12 h

after transfection. Luciferase assay was performed 24 h after hypoxia treat-ment to reflect promoter activity. pCMV-Rluc was cotransfected to normalizetransfection efficiency. The numbers represent mean � SEM; n � 4; *, P �0.0001 by Student’s t test. (I) Mutation in the HRE site abolishes the effect ofhypoxia on BACE1 promoter activity. The HRE site in the BACE1 promoter ofpB1P-H was mutated by site-directed mutagenesis. The mutant construct wastransfected into SH-SY5Y cells and exposed to 2% O2 (hypoxia) or 21% O2

(normoxia) for 48 h. (J) Swedish APP stable SH-SY5Y cells were exposed to 2%or 21% O2 for 24 h before RNA extraction. (K) Hypoxia increased BACE1 mRNAby �1.5 times. *, P � 0.05 by Student’s t test.

18728 � www.pnas.org�cgi�doi�10.1073�pnas.0606298103 Sun et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 11

, 202

1

pHIF-1� plasmid showed robust HIF-1� expression (Fig. 1E).Overexpression of HIF-1� had no effect on luciferase activity incells transfected with pB1P-I (P � 0.05). In contrast, luciferaseactivity increased dramatically in cells cotransfected with pB1P-Hand HIF-1�, showing a 4.63 � 0.50-fold increase compared with thenon-HIF-1� transfected cells (P � 0.001) (Fig. 1F). HIF-1� alsosignificantly increased luciferase activity in cells transfected withpositive control plasmid pEpoE-Luc (15.81 � 0.23-fold, P � 0.001,relative to vector control) (Fig. 1F). To examine whether up-regulation of human BACE1 gene transcription by hypoxia wasmediated by transcription factor HIF-1, an HIF-1� siRNA assaywas performed to knock down HIF-1� expression. The HIF-1�siRNA significantly inhibited HIF-1� expression (Fig. 1G). Incontrol siRNA transfected cells, hypoxia treatment markedly in-creased luciferase activity in the cells transfected with humanBACE1 promoter plasmid pB1P-H or with positive control plasmidpEpoE-Luc. In contrast, hypoxia treatment had no effect on thepromoter activity of human BACE1 promoter plasmids pB1P-Hand pEpoE-Luc after transfection of HIF-1� siRNA (Fig. 1H).These data indicate that knock-down of HIF-1� expression blockedthe effect of hypoxia on the human BACE1 gene transcription. Tofurther confirm that HIF-1 can bind specifically to this HRE in theBACE1 promoter, we generated a mutant BACE1 plasmid, pB1P-HIFm. The putative HRE core binding sequence (5�aCGTg) in theBACE1 promoter plasmid pB1P-H was changed to 5�-aAAAg bysite-directed mutagenesis, creating mutant plasmid pB1P-HIFm,which lacks the HIF-1 binding site. Hypoxia treatment had no effecton the promoter activity of pB1P-HIFm in SH-SY5Y cells (79.26 �8.52% under hypoxia vs. 100 � 16.25% under normoxia; P � 0.05)(Fig. 1I). The mutation resulted in an inability of the BACE1 genepromoter to respond under hypoxic conditions. Taken together,these data demonstrate that the BACE1 gene promoter contains ahypoxia-inducible enhancer and that the HRE site is physiologicallyfunctional in transcriptional regulation of BACE1 gene expression.

To further investigate whether hypoxia plays an important role

in the transcription of human BACE1, we analyzed endogenousBACE1 mRNA levels by using quantitative RT-PCR, with�-actin transcriptional levels serving as the internal control.Under hypoxic conditions, endogenous BACE1 mRNA levels inSH-SY5Y cells were increased to 144.8 � 10.42% relative tonormoxia (P � 0.05) (Fig. 1 J and K). The data are consistentwith the BACE1 promoter assay results, indicating that hypoxiaplays an important role in transcriptional regulation of thehuman BACE1 gene.

Hypoxia Increases APP Processing and A� Generation by Up-Regulat-ing BACE1 Activity. A� is derived from APP by �- and �-secretasecleavages and is the major component of neuritic plaques.BACE1 is the major �-secretase in vivo. To determine whetherup-regulation of BACE1 gene expression under hypoxic condi-tions affects APP processing at the �-secretase site to generateA�, we analyzed the protein levels of BACE1, APP C99, and A�in SH-SY5Y cells that stably overexpress Swedish mutantAPP695. BACE1 protein and C99, the major �-secretase cleav-age product of APP, were markedly increased under hypoxicconditions (Fig. 2A). Quantification showed that C99 levels wereincreased by 221.7 � 7.32% and 302.1 � 1.74% (P � 0.001) (Fig.2B) and that BACE1 protein levels were increased to 143.4 �1.19% and 304.5 � 1.69% after 12 h and 24 h of hypoxiatreatment, respectively (P � 0.001) (Fig. 2C). The production ofA�40 and A�42 was significantly increased to 128.6 � 0.56% and159.8 � 1.17%, respectively, in hypoxia treated cells relative tocontrol cells (P � 0.0001) (Fig. 2 D and E). To further confirmthat hypoxia has a similar effect on wild-type (WT) APPprocessing, HAW1 cells that stably express a WT APP695protein were treated with hypoxia. The �-secretase cleavageproducts, the C-terminal fragment � (CTF�) including C99 andC89, were significantly increased under hypoxic conditions (Fig.2F), and quantification showed that the CTF� levels wereincreased by 148.97 � 8.89% (P � 0.001 relative to control) (Fig.

Fig. 2. Hypoxia increases A� production by up-regulating BACE1 activity. (A) SH-SY5Y cells stably overexpressing Swedish mutant APP were exposed to 2% O2

for 0, 12, or 24 h. C20 antibody was used to detect APP CTFs; 208 antibody was used to detect BACE1. (B and C) Quantification of C99 (B) and BACE1 protein (C)levels. *, P � 0.001 by ANOVA. (D and E) ELISA was performed to measure A�40 (D) and A�42 (E) in culture media from SH-SY5Y cells stably overexpressing Swedishmutant APP exposed to 2% O2 (hypoxia) or 21% O2 (normoxia) for 24 h. *, P � 0.0001 by Student’s t test. (F) WT APP695 cell HAW1 was cultured under 2% O2

for 12 h, and APP CTFs were detected with C20 antibody. (G) The levels of CTF� including C99 and C89 were quantitated. *, P � 0.001 by Student’s t test. (H andI) The conditioned media from the same hypoxia-treated cells were analyzed for A�40 (H) and A�42 (I) levels. The numbers represent mean � SEM; n � 3; *, P �0.0001 by Student’s t test.

Sun et al. PNAS � December 5, 2006 � vol. 103 � no. 49 � 18729

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Nov

embe

r 11

, 202

1

2G). Hypoxia also markedly increased A�40 and A�42 levels inHAW1 cells to 252.6 � 6.52% and 204.1 � 0.96%, respectively(P � 0.0001 relative to controls) (Fig. 2 H and I). These dataindicate that hypoxia can elevate BACE1 protein expression andin turn increase APP processing at the �-secretase site topotentiate A� generation.

Hypoxia Facilitates �-Secretase Cleavage of APP and A� Deposition inAPP23 Transgenic Mice. Our data indicate that hypoxia facilitatesBACE1 gene transcription and increases APP processing at the�-secretase site to generate A�. To investigate whether hypoxiawould facilitate AD pathogenesis in vivo, APP23 transgenicmice, a valued AD mouse model, were subjected to hypoxiatreatment. APP23 mice carry the human Swedish mutantAPP751 transgene driven by the Thy1.2 promoter element,which drives the transgene to be specifically expressed in neurons(35). APP23 mice were subjected to 8% O2 for 16 h/day for 1month; age-matched control APP23 mice were exposed tonormal air (21% O2). To determine whether hypoxia affects

APP processing in vivo, the levels of APP CTFs in APP23 braintissue lysates were assayed using Western blot analysis (Fig. 3A).C99, the major �-secretase product, was significantly increasedin the hypoxic mice by 151.1 � 4.15% relative to controls (P �0.01) (Fig. 3B). To examine the effect of hypoxia on A�generation in the transgenic mice, A� ELISA analysis wasperformed. Hypoxia treatment increased the levels of A�40generation by 357.5 � 137.01% relative to controls (P � 0.001)and increased the levels of A�42 by 184.6 � 79.08% (P � 0.001)(Fig. 3 C and D). These data demonstrate that hypoxia up-regulates �-secretase cleavage of APP and thus A� production.APP23 mice develop amyloid plaques in neocortex and hip-pocampus as early as 6 months of age and progressively with age(35). To investigate whether the altered APP processing causedby hypoxia resulted in A� deposition and an increase in neuriticplaque formation in vivo, the hypoxia-treated and control APP23mice were killed and 4G8 immunostaining (Fig. 3E) and thio-flavin S staining (Fig. 3G) were used to detect A�-containingneuritic plaques in brain. Neuritic plaque formation was signif-

Fig. 3. Hypoxia increases �-secretase cleavage of APP and A� deposition in APP23 transgenic mice. Eight-month-old APP23 mice were treated with 8% O2

(hypoxia) for 16 h/day for 1 month. Ten mice per group were used for hypoxia and normoxia treatment. (A) Half brains from hypoxic and age-matched normoxiccontrol mice were lysed in RIPA-Doc lysis buffer and separated with 16% Tris-Tricine SDS/PAGE gel. C99 was detected by C20 polyclonal antibody. �-actin wasdetected by anti-�-actin antibody AC-15 as the internal control. (B) Quantification showed that C99 was significantly increased in hypoxia-treated mice. *, P �0.01 by Student’s t test. (C and D) ELISA was performed to measure A�40 (C) and A�42 (D) levels in mouse brain tissue lysates. *, P � 0.001 by Student’s t test.(E) 4G8 immunostaining. The other half brains were dissected from hypoxic and age-matched normoxic control mice, fixed, and sectioned. Neuritic plaques weredetected by A�-specific mAb 4G8 (Signet Laboratories) and the DAB method. The plaques were visualized under a microscope at 40 magnification. Moreneuritic plaques were stained in hypoxic mice (c and d) compared with age-matched control mice (a and b). Arrows point to plaques. Magnification at 200reveals more and larger plaques in hypoxic mice ( f) than in normoxic mice (e). (F) Quantification of neuritic plaques. *, P � 0.001 by ANOVA. (G) Thioflavin Sstaining. Neuritic plaques were further confirmed by using thioflavin S fluorescent staining and were visualized under a microscope at 100 magnification. aand c are sections of frontal cortex; b and d are sections of hippocampus. More neuritic plaques were seen in hypoxia-treated mice (c and d) than in age-matchedcontrols (a and b). Arrows point to green fluorescent neuritic plaques. (H) A mouse BACE1 promoter was cloned into the pGL3-Basic vector and transfected intocells. Hypoxia increased luciferase activity in pB1Pm-A transfected cells that contained two putative HRE consensus sites. *, P � 0.01 by Student’s t test. (I) Gelshift assay demonstrated that the mouse BACE1 promoter also contains an HIF-1 transcription factor binding site. Gel shift assay was performed using a32P-labeled [32P]mBACE1-HRE probe in which the two putative HRE consensus sites were juxtaposed. Lane 1, probe alone without nuclear extract; lane 2,incubation of HeLa nuclear extract with probe retarded the migration of free probes to form a new HIF-1-DNA complex; lanes 3 and 4, competition assays adding10- and 100-fold molar excess of Epo-HRE consensus oligonucleotides. (J) Endogenous BACE1 mRNA level in the WT control mice. The control mice were subjectedto the hypoxia treatment, and total RNA samples were extracted from the mouse brains for RT-PCR assay. �-actin was amplified as the internal control. (K) Hypoxiasignificantly increased endogenous BACE1 mRNA in WT mice by �1.5 times. *, P � 0.0001 relative to normoxia WT mice controls by Student’s t test.

18730 � www.pnas.org�cgi�doi�10.1073�pnas.0606298103 Sun et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 11

, 202

1

icantly increased in APP23 mice under hypoxic conditionsrelative to normoxic controls (Fig. 3 E and G). More and largerplaques were seen in hypoxic mice than in normoxic mice (Fig.3E). Quantification showed that hypoxia-treated APP23 micehad �1.5-fold more plaques than normoxic mice (2.8 � 0.5 vs.1.8 � 0.4 per slice; P � 0.001) (Fig. 3F).

To confirm that hypoxia facilitates APP processing and A�generation in the transgenic mice by up-regulating mouse BACE1gene transcription, we cloned the mouse BACE1 promoter. Se-quence analysis reveals that there are two putative HRE consensussites in the mouse BACE1 5�UTR. Hypoxia markedly increasedluciferase activity by 199.81 � 8.97% in cells transfected with mousepromoter plasmid pB1Pm-A, which contains the HRE sites (P �0.01) (Fig. 3H). Gel shift assay showed that the mouse BACE1promoter region contains HRE (Fig. 3I). To determine whetherhypoxia up-regulates mouse BACE1 gene transcription in vivo, WTcontrol mice were subjected to hypoxia treatment and the mouseendogenous BACE1 mRNA levels were measured using quantita-tive RT-PCR. Under hypoxic conditions, endogenous BACE1mRNA levels in the WT mice were significantly increased to156.7 � 2.45% relative to normoxia (P � 0.0001) (Fig. 3 J and K).These results suggest that mouse BACE1 gene expression, likehuman BACE1 gene expression, can be up-regulated by hypoxia atthe transcriptional level.

Hypoxia Potentiates the Memory Deficit in APP23 Mice. To furtherinvestigate whether hypoxia affects learning and memory in ADpathogenesis, behavioral tests were performed after APP23 miceunderwent 1 month of hypoxia treatment. The Morris watermaze was used to determine the effect of hypoxia on spatialmemory. In visible platform tests, the hypoxic and normoxicAPP23 mice had a similar escape latency (37.15 � 2.58 s and30.90 � 2.60 s, respectively; P � 0.05) (Fig. 4A) and path length(6.52 � 0.66 m and 6.38 � 1.07 m, respectively; P � 0.05) (Fig.4B), indicating that hypoxia did not affect mouse motility orvision. In the hidden-platform swimming test, hypoxia-treatedAPP23 mice showed a longer latency (P � 0.005; Fig. 4C) andswam a significantly longer distance (P � 0.001; Fig. 4D) to reachthe platform compared with normoxic mice. Furthermore, hy-poxic APP23 mice had significantly fewer platform-passing timesin the probe trial (2.5 � 0.48 s vs. 4.29 � 0.52 s; P � 0.05) (Fig.2E). These data demonstrate that hypoxia significantly poten-tiates the memory deficit in APP23 mice.

Hypoxia is one of the major common components of vascularrisk factors for AD pathogenesis such as stroke, hypertension,atherosclerosis, and diabetes. Hypoxia can also arise fromcerebral amyloid angiopathy. Previous studies (32, 36) showedthat BACE1 expression and enzymatic activity are elevated inthe brains of sporadic AD patients, and the degree of elevationis correlated with A� production. The mechanism by whichvascular factors facilitate AD neurodegeneration in AD patho-genesis has not yet been defined. It has been reported that HIF-1is increased in the brains of elderly people (16). By subjecting ADtransgenic model mice to low-oxygen conditions, we clearlydemonstrated that hypoxia facilitates plaque formation andenhances memory deficits. Our study shows that BACE1 is oneof the downstream target genes of the hypoxia signaling pathwayand that hypoxia potentiates APP processing to generate A� byactivating BACE1 transcription through an HRE in its promoterregion, leading to increased BACE1 expression. A slight increasein BACE1 expression could lead to a dramatic increase in A�production (37). HIF-1� can also be up-regulated by A� (38).Because our data showed that hypoxia increases A� production,a positive feedback circuit may dramatically increase A� pro-duction and deposition in AD pathogenesis. It has been reportedthat chronic hypoxic insults could alter APP processing at the�-secretase site, resulting in reduction of C83 and sAPP�(39–41). Although we have not observed a significant decrease

in the �-secretase pathway, our data also show that chronichypoxia can alter APP processing by its effects on the �-secretasepathway in vivo. Furthermore, hypoxia can trigger the transcrip-tion of many other genes besides BACE1. Prolonged or severehypoxia can also trigger apoptosis in the brain, which maycontribute to the neuronal loss and memory impairment seen inAD patients. Transient hypoxic injury to cortical neurons cancause the mitochondrial dysfunction, impaired membrane integ-rity, and altered APP processing that occur in AD (42). Ergoloidmesylates (e.g., Hydergine) have been the longest used putativecognitive enhancement drugs and are one of the classes of drugsused in AD treatment. As a mild cerebral vasodilator, Hyderginecan decrease hypoxia in the early stages of AD and improvesymptoms in AD patients (43). Our study suggests that enhanc-ing oxygen supply by improving cerebral perfusion or by usingvasodilators may have pharmaceutical potential for treatingcognitive impairments, thereby benefiting AD patients.

MethodsTransgenic Mice and Hypoxia Treatment. APP23 transgenic micecarry the human APP751 cDNA with the Swedish doublemutation at positions 670/671 (KM3NL) under control of themurine Thy-1.2 expression cassette (35). Littermates of APP23mice carrying no human Swedish mutant APP751 cDNA wereused as WT controls. APP23 mice (8 months of age, n � 20, 50%

Fig. 4. Hypoxia potentiates the memory deficit in Swedish mutant APP trans-genic mice. A Morris water maze protocol with 1 day of visible platform tests and4daysofhiddenplatformtests,plusaprobetrialonday6.Animalmovementwastracked and recorded by an HVS 2020 Plus image analyzer (HVS Image, Hampton,UK). Ten mice per group were used for hypoxia and normoxia treatment. (A)During the first day of visible platform tests, mice were trained in five contiguoustrials. The hypoxic and normoxic APP23 mice exhibited a similar latency to escapeonto the visible platform. P � 0.05 by Student’s t test. (B) Hypoxic and normoxicAPP23 mice swam a similar distance before escaping onto the visible platform inthe visible platform test. P � 0.05 by Student’s t test. (C) In hidden platform tests,mice were trained with six trials per day for 4 days. Hypoxic APP23 mice showeda longer latency to escape onto the hidden platform. P � 0.005 by two-wayANOVA. (D) Hypoxic APP23 mice swam farther before escaping onto the hiddenplatform. P � 0.001 by two-way ANOVA. (E) In the probe trial on day 6, hypoxicAPP23 mice crossed the target platform significantly fewer times than controls. *,P � 0.05 by Student’s t test.

Sun et al. PNAS � December 5, 2006 � vol. 103 � no. 49 � 18731

MED

ICA

LSC

IEN

CES

Dow

nloa

ded

by g

uest

on

Nov

embe

r 11

, 202

1

female) were assigned randomly to hypoxia and control groups.The hypoxia group was treated in a hypoxia chamber at 8% O2

for 16 h/day for 1 month. The oxygen level was regulated byinfusing nitrogen into a semisealable chamber controlled by aPROOX model 110 controller (BioSpherix , Redfield, NY).

Morris Water Maze. The water maze test was performed in a pool1.5 m in diameter. In the hidden platform trials, a 10-cm-diameter platform was placed in the southeastern quadrant ofthe pool. The procedure consists of 1 day of visible platform testsand 4 days of hidden platform tests, plus a probe trial 24 h afterthe last hidden platform test. In the visible platform test, micewere tested for five contiguous trials, with an intertrial intervalof 30 min. In the hidden platform tests, mice were trained for sixtrials, with an intertrial interval of 1 h. Tracking of animalmovement was achieved with an HVS 2020 Plus image analyzer(HVS Image). Data were analyzed by two-way ANOVA.

Neuritic Plaque Staining. Mice were killed after the behavioral tests,and the half brains were fixed and sectioned with a Leica (Deerfield,IL) cryostat to 40 �m thickness. Every sixth slice with the samereference position of the mice was mounted onto the slides forstaining. The slices were immunostained with biotinylated mono-clonal 4G8 antibody (Signet Laboratories, Dedham, MA) at 1:500dilution. Approximately 24 slices were stained for each mouse.Plaques were visualized by the avidin-biotin-peroxidase complex(ABC) and diaminobenzidine (DAB) method and counted undermicroscopy with 40 magnification. Plaques were quantitated byaverage plaque count per slice for each mouse. The data wereanalyzed by two-way ANOVA. Thioflavin S staining of plaques was

performed with 1% thioflavin S, and the green fluorescence-stained plaques were visualized with fluorescent microscopy.

Cell Culture, Transfection, Hypoxia Treatment, and Luciferase Assay.SH-SY5Y cells, HEK293 cells, and N2A cells were grown incomplete DMEM. All cells were maintained at 37°C in an incubatorcontaining 5% CO2. Cells were transfected with plasmid DNA byusing Lipofectamine 2000 (Invitrogen, Carlsbad, CA). A Swedishmutant APP stable cell line was generated by transfecting pZ-APPsw into SH-SY5Y cells and selected by 1,000 �g/ml Zeocine(Invitrogen). HAW1 cells are the WT APP695 stably expressingHEK293 cells. Hypoxia treatments were performed by incubatingcells in a 37°C incubator containing 2% O2 and 5% CO2. The Renilla(sea pansy) luciferase vector pCMV-Rluc was cotransfected tonormalize transfection efficiency. Luciferase assay was performed48 h after transfection by using a dual-luciferase reporter assaysystem (Promega, Madison, WI) as described previously (17).

Plasmids, Gel Shift Assay, Immunoblotting, siRNA Assay, QuantitativeRT-PCR, and A�40/42 Sandwich ELISA. Details of these proceduresare provided in Supporting Methods, which is published assupporting information on the PNAS web site.

We thank Brain MacLean, Xiaojian Sun, and Jane Wang for technicalassistance, and Anthony G. Phillips and Kelley Bromley for helpful discus-sion. This work was supported by the Canadian Institutes of HealthResearch (CIHR), The Jack Brown and Family Alzheimer’s ResearchFoundation, and The Michael Smith Foundation for Health Research(W.S.). W.S. is the holder of The Canada Research Chair in Alzheimer’sDisease. W.Z. was the recipient of The Arthur and June Willms Fellowships.G.H. was supported by The Chinese Scholarship Council Award. F.D. wasthe recipient of a Natural Sciences and Engineering Research Council(NSERC) graduate fellowship.

1. Mattson MP (2004) Nature 430:631–639.2. White L, Petrovitch H, Hardman J, Nelson J, Davis DG, Ross GW, Masaki K,

Launer L, Markesbery WR (2002) Ann NY Acad Sci 977:9–23.3. Snowdon DA, Greiner LH, Mortimer JA, Riley KP, Greiner PA, Markesbery

WR (1997) J Am Med Assoc 277:813–817.4. Altieri M, Di Piero V, Pasquini M, Gasparini M, Vanacore N, Vicenzini E,

Lenzi GL (2004) Neurology 62:2193–2197.5. Schneider JA, Wilson RS, Cochran EJ, Bienias JL, Arnold SE, Evans DA,

Bennett DA (2003) Neurology 60:1082–1088.6. Vermeer SE, Prins ND, den Heijer T, Hofman A, Koudstaal PJ, Breteler MM

(2003) N Engl J Med 348:1215–1222.7. Honig LS, Kukull W, Mayeux R (2005) Neurology 64:494–500.8. Heyman A, Fillenbaum GG, Welsh-Bohmer KA, Gearing M, Mirra SS, Mohs

RC, Peterson BL, Pieper CF (1998) Neurology 51:159–162.9. Huang LE, Willmore WG, Gu J, Goldberg MA, Bunn HF (1999) J Biol Chem

274:9038–9044.10. Huang LE, Bunn HF (2003) J Biol Chem 278:19575–19578.11. Pugh CW, Ratcliffe PJ (2003) Nat Med 9:677–684.12. Halterman MW, Miller CC, Federoff HJ (1999) J Neurosci 19:6818–6824.13. Pulsinelli WA, Brierley JB, Plum F (1982) Ann Neurol 11:491–498.14. Chen J, Nagayama T, Jin K, Stetler RA, Zhu RL, Graham SH, Simon RP

(1998) J Neurosci 18:4914–4928.15. Sharp FR, Bernaudin M (2004) Nat Rev Neurosci 5:437–448.16. Lu T, Pan Y, Kao SY, Li C, Kohane I, Chan J, Yankner BA (2004) Nature

429:883–891.17. Sun X, Wang Y, Qing H, Christensen MA, Liu Y, Zhou W, Tong Y, Xiao C,

Huang Y, Zhang S, et al. (2005) FASEB J 19:739–749.18. Sun X, He G, Song W (2006) FASEB J 20:1369–1376.19. von Arnim CAF, Kinoshita A, Peltan ID, Tangredi MM, Herl L, Lee BM,

Spoelgen R, Hshieh TT, Ranganathan S, Battey FD, et al. (2005) J Biol Chem280:17777–17785.

20. Li Q, Sudhof TC (2004) J Biol Chem 279:10542–10550.21. Pastorino L, Ikin AF, Lamprianou S, Vacaresse N, Revelli JP, Platt K, Paganetti

P, Mathews PM, Harroch S, Buxbaum JD (2004) Mol Cell Neurosci 25:642–649.22. Kitazume S, Tachida Y, Oka R, Shirotani K, Saido TC, Hashimoto Y (2001)

Proc Natl Acad Sci USA 98:13554–13559.

23. Lichtenthaler SF, Dominguez DI, Westmeyer GG, Reiss K, Haass C, Saftig P,De Strooper B, Seed B (2003) J Biol Chem 278:48713–48719.

24. He W, Lu Y, Qahwash I, Hu XY, Chang A, Yan R (2004) Nat Med 10:959–965.25. Ge YW, Maloney B, Sambamurti K, Lahiri DK (2004) FASEB J 18:1037–1039.26. Zhou W, Song W (2006) Mol Cell Biol 26:3353–3364.27. Christensen MA, Zhou W, Qing H, Lehman A, Philipsen S, Song W (2004) Mol

Cell Biol 24:865–874.28. Fukumoto H, Cheung BS, Hyman BT, Irizarry MC (2002) Arch Neurol

59:1381–1389.29. Fukumoto H, Rosene DL, Moss MB, Raju S, Hyman BT, Irizarry MC (2004)

Am J Pathol 164:719–725.30. Russo C, Schettini G, Saido TC, Hulette C, Lippa C, Lannfelt L, Ghetti B,

Gambetti P, Tabaton M, Teller JK (2000) Nature 405:531–532.31. Holsinger RM, McLean CA, Beyreuther K, Masters CL, Evin G (2002) Ann

Neurol 51:783–786.32. Yang LB, Lindholm K, Yan R, Citron M, Xia W, Yang XL, Beach T, Sue L,

Wong P, Price D, et al. (2003) Nat Med 9:3–4.33. Sun X, Tong Y, Qing H, Chen CH, Song W (2006) FASEB J 20:1361–1368.34. Madan A, Curtin PT (1993) Proc Natl Acad Sci USA 90:3928–3932.35. Sturchler-Pierrat C, Abramowski D, Duke M, Wiederhold KH, Mistl C,

Rothacher S, Ledermann B, Burki K, Frey P, Paganetti PA, et al. (1997) ProcNatl Acad Sci USA 94:13287–13292.

36. Li R, Lindholm K, Yang LB, Yue X, Citron M, Yan R, Beach T, Sue L, SabbaghM, Cai H, et al. (2004) Proc Natl Acad Sci USA 101:3632–3637.

37. Li Y, Zhou W, Tong Y, He G, Song W (2006) FASEB J 20:285–292.38. Soucek T, Cumming R, Dargusch R, Maher P, Schubert D (2003) Neuron

39:43–56.39. Auerbach ID, Vinters HV (2006) J Neuropathol Exp Neurol 65:610–620.40. Webster NJ, Green KN, Peers C, Vaughan PF (2002) J Neurochem 83:1262–

1271.41. Webster NJ, Green KN, Settle VJ, Peers C, Vaughan PF (2004) Brain Res Mol

Brain Res 130:161–169.42. Chen GJ, Xu J, Lahousse SA, Caggiano NL, de la Monte SM (2003) J

Alzheimers Dis 5:209–228.43. Olin J, Schneider L, Novit A, Luczak S (2001) Cochrane Database Syst Rev

CD000359.

18732 � www.pnas.org�cgi�doi�10.1073�pnas.0606298103 Sun et al.

Dow

nloa

ded

by g

uest

on

Nov

embe

r 11

, 202

1